2. L.D. Miranda et al. / Journal of Membrane Science 344 (2009) 26–31 27
2. Experimental methods
2.1. Anodic aluminum oxide membrane
An aluminum strip (0.5 mm thick, purity 99.999%) was
degreased in acetone and partly covered with electroplating tape.
Exposed aluminum was electropolished and anodized using the
two-step process of Masuda and Satoh [15]. The initial aluminum
anodization step was conducted at 40 V in 0.3 M oxalic acid at 10 ◦C.
After 17 h the oxide layer was removed using a mixture of phos-
phoric acid (6 wt%) and chromic acid (1.8 wt%) over a period of 1 h
at 60 ◦C. The aluminum anodization was then repeated for 10 h.
A protective polymer coat was applied to the oxide surface, and
the remaining aluminum was removed using a saturated mercuric
chloride solution. The aluminum oxide membrane was subse-
quently etched in 5% phosphoric acid (45 min at 30 ◦C). Mechanical
removal of the polymer coating produced a freestanding AAO mem-
brane. The AAO membrane was then heated in air between two
quartz plates (30 min at 900 ◦C) [16] causing a phase transition from
amorphous to gamma alumina [17].
2.2. Carbon nanotube synthesis
In the first stage of carbon nanotube synthesis, the AAO mem-
brane was placed in a quartz tube furnace, edge-up in a quartz
boat. The furnace-tube was purged with argon and the temper-
ature was raised to 750 ◦C. After thermal stabilization, the argon
flow was maintained at 10.5 mL/min and ethylene was added to
the gas stream at 0.35 mL/min. Carbon deposition was allowed for
16 h, after which the flow of ethylene was terminated and the fur-
nace was cooled to room temperature with a continuing flow of
argon.
2.3. Preparation of the CNT membrane
The chemical vapor deposition process generated carbon nan-
otubes within the pores of the AAO membrane and deposited a thin
carbon film on the surface of the membrane. The carbon film was
removed (ion milled, GATAN 691) from both sides of the membrane,
producing a smooth surface. The open pores of the CNT membrane
produced in this process are shown in Fig. 1.
A membrane module was then created using vacuum epoxy to
mount the membrane on a stainless steel frit (Fig. 2). To demon-
strate that the vacuum epoxy could eliminate leaks around the edge
of the membrane, an extensive carbon deposition process was used
Fig. 1. Scanning electron microscope (Hitachi S-4800 Field Emission) image of the
CNT membrane surface after ion milling. The dark areas in the image are open pores,
the light rim around the pore is the CNT and the remainder substrate is the aluminum
oxide film.
Fig. 2. Schematic of the CNT membrane module.
to seal the pores of an AAO membrane with a thick carbon film. Mass
spectrometry results showed insignificant gas introduction around
the blocked membrane, ensuring that ion currents obtained with
the CNT membrane would be attributable to diffusion through the
CNT pores (results not shown).
2.4. Mass spectrometry experiments
The selectivity of the CNT membrane was examined using two
Airgas certified gas mixtures: Gas Mixture A was composed of
0.50% methane (CH4), 1.50% argon (Ar), 0.20% carbon dioxide (CO2),
10.00% oxygen (O2) and 87.80% nitrogen (N2). Gas Mixture B was
composed of 1.01% CH4, 2.04% Ar, 2.03% CO2, 14.96% O2, and 79.96%
N2. Mass spectrometry experiments were conducted using the sys-
tem shown in Fig. 3. The two-position stage rotary valve (Valco
Instruments Co. Inc.) in the center of Fig. 3 had four connections
whereby the inlet of the MS (Inficon, Transpector 2.0 Gas Ana-
lyzer System) could be connected directly to the CNT membrane
module or the CT (Restek, Hydrogard FS, 0.1 mm ID) without break-
ing vacuum. A fourth valve connection provided coupling to a
diaphragm pump. This diaphragm pump was used to reduce the
pressure within the CNT membrane module and the CT when
these inlets were not in use. This precluded sudden increases in
MS vacuum chamber pressure when the valve was switched to
either inlet. Experiments were begun after a steady baseline sig-
nal was observed. Gases were analyzed by their mass-to-charge
(m/z) ratio. The m/z signals of CH4, N2, O2, Ar, and CO2 were ana-
lyzed at 15, 28, 32, 40, and 44, respectively, using a Faraday cup
detector. Gas mixtures were delivered to either the CNT membrane
or the CT through a series of Swagelok valves. For direct compar-
ison of the CNT membrane and the CT introduction systems, inlet
flow rates were matched. This created similar total pressures inside
the MS ionization region and, thereby, similar ionization condi-
tions. The gas exhaust of both the CNT membrane module and the
CT setup was connected to a single exhaust line where flow rate
and pressure could be observed and controlled. Experimental runs
were performed at a steady flow rate over a range of pressures.
Total pressure inside the ionization region was measured using the
pressure-reading software of the Transpector, and ranged between
3.3 × 10−3 and 6.7 × 10−3 Pa.
3. Results and discussion
3.1. CNT membrane
The two-step anodization process developed by Masuda and
Satoh [15] creates a uniform and monodisperse array of pores.
The AAO membrane produced in this work was 70 m thick and
had channels 50 nm in diameter. The ethylene gas pyrolyzed into
a graphitic structure on the alumina surface [9]. CNTs within the
3. 28 L.D. Miranda et al. / Journal of Membrane Science 344 (2009) 26–31
Fig. 3. Schematic drawing of the mass spectrometry experimental setup.
membrane had an average wall thickness of 5 nm, and outside
diameters and lengths identical to those of the AAO pores. Slow
flow rates of ethylene at concentrations less than 3.5%, and lengthy
periods of deposition at 750 ◦C, produced uniform growth of CNT
walls. In contrast, rapid carbon deposition thickens the surface film
of carbon and obstructs CNT wall growth.
The AAO membrane has remarkable strength properties [18],
and AAO membrane strength can be improved by controlling
pore-cell dimension and crystalline structure through calcination
[17,19]. In addition, support by a stainless steel frit in the membrane
module reduced stress on the membrane and increased membrane
tolerance to high pressure differentials between the sample and the
mass spectrometer vacuum.
3.2. Raman spectrum of the CNT membrane
Raman spectra of carbon allotropes revealed the characteris-
tic crystal structure of the nanotubes. The extent of sp2 and sp3
bonding on CNTs produces a unique Raman fingerprint [20]. Fig. 4a
shows a Raman spectrum for the synthesized CNTs. All spectra
showed first-order Raman bands at ∼1350 cm−1 (D band) and
∼1580 cm−1 (G band). The D band to G band intensity ratio (ID/IG)
is linearly related to the degree of CNT crystallinity [21]. The ID/IG
value of 0.80 shown in Fig. 4a indicates a low degree of crystallinity
compared to highly oriented pyrolytic graphite. Raman spectra
and transmission electron microscope (TEM, Hitachi 7100) imagery
(Fig. 4b) of our CNTs confirmed a turbostratic structure [22–24].
3.3. Mass spectrometry
Fig. 5 shows MS ion currents produced by methane (˚CH4
) using
(a) capillary tube introduction and (b) CNT membrane introduc-
tion. Each horizontal section in Fig. 5 was produced during steady
state flow at constant pressure. The different section heights in
Fig. 5 were produced by successive pressure increases in the sys-
tem. Observations of the ion currents (˚G) produced by each gas
(G) were obtained in triplicate.
Gases that passed through the inlet systems were constantly
evacuated by the pumps achieving a steady state condition and
thus constant pressure inside the vacuum chamber. ˚G is propor-
tional to the partial pressure (PG) of each gas in the MS vacuum
chamber via the relationship ˚G = S·PG, where S is the MS sen-
sitivity factor. Then ˚G is proportional to the quantity of a gas
that passes through the inlet plane in a known amount of time.
Therefore, throughput (QG) of the inlet system is proportional to
˚G times a proportionality factor k, QG = k·˚G. The properties of
each gas and each introduction system were examined by averag-
ing the measured values of ˚G. The conductance (CG) of each gas
was measured using the following equation: CG = QG/ P where P
is the total pressure gradient. CG was normalized by the mole frac-
tion ( ) of each gas via the relationship C = CG/ . The normalized
conductance (C ) was then plotted against the total pressure on the
inlet system (Fig. 6). The results in Fig. 6 show distinct differences
in the transmission characteristics of the two introduction systems
and distinct differences for different gases.
C values in the CNT introduction system uniformly decrease
with increasing total pressure. The extent of this decrease is
greatest for CH4 and becomes progressively smaller in the order
CH4> N2 > O2 > Ar > CO2. The influence of total pressure on trans-
mission of CO2 in the CNT system is quite small. In contrast, C
in the CT system generally has an inverse order. Pressure effects
are smallest for CH4 and, for the remaining gases, the pressure-
dependent slopes are progressively less negative (N2 to O2) and
then increasingly positive (Ar to CO2).
Fig. 4. Results from analysis of CNT membrane: (a) Raman spectrum of the CNT
membrane and (b) TEM image of a CNT.
4. L.D. Miranda et al. / Journal of Membrane Science 344 (2009) 26–31 29
Fig. 5. Ion currents produced by CH4 using (a) capillary tube and (b) CNT membrane
introduction systems.
Graph b of Fig. 6 shows that N2 from Gas Mixture B has a some-
what less negative slope than the corresponding N2 signal of Gas
Mixture A. This is due to the increased concentration of CO2 in the
gas mixture, thus contributing CO+ fragment ions to the m/z 28
ion current signal. Nonetheless, the CNT and CT conductance trend
for N2 is real as this gas has the highest concentration in both gas
mixtures.
Fig. 6 shows, for each component gas, that CNT membranes
have relatively simple flux characteristics. Increasing total pressure
decreases the conductance of each gas. This simple behavior is not
observed in the CT introduction system, in which gas conductance
both decreases (N2) and increases (O2, Ar and CO2) with increasing
pressure. In each plot the dependence of normalized conductance
on pressure is the same for Gas Mixtures A and B. However, in each
case, normalized conductances are slightly higher for Gas Mixture
A. This offset arises because the ion currents (˚G) used to calculate
normalized conductance for each gas were not background cor-
rected for contributions of residual gas in the vacuum housing. The
experiments for each gas mixture were performed on different days
and thus a slight difference in the partial pressure of each residual
gas led to the observed offsets.
In Fig. 7 CNT and CT gas transmission characteristics as a function
of pressure are compared in the following form: C (CNT)/C (CT).
Fig. 7 shows that, relative to the CT system, the CNT membrane
is selectively transmissive to CH4 over a range of conditions. In
the case of N2, the CNT and CT introduction systems are generally
comparable (C (CNT)/C (CT)∼1) over a range of pressures. For the
remaining gases, O2, Ar and CO2, the CNT membrane is less conduc-
tive than the CT system, and the magnitudes of the differences in
transmission become larger with increasing pressure. The results
shown in Fig. 7 are in general agreement with previous compar-
isons of CNT gas transmission properties obtained through analysis
of discrete samples [14].
The synthesized CNT membranes have a low degree of crys-
tallinity (Fig. 4b), a high surface area, and thereby a high propensity
for adsorption [25]. The adsorption of gases to the surfaces of
nanotube-channels causes a temperature-dependent [26,27] and
Fig. 6. Normalized conductance (C ) against total pressure for Gas Mixtures A and
B using capillary tube and CNT membrane introduction system.
5. 30 L.D. Miranda et al. / Journal of Membrane Science 344 (2009) 26–31
Fig. 7. CNT membrane and capillary tube conductance ratios over the total pressure.
Fig. 8. CNT normalized transport resistance of each component gas with increasing
pressure.
pressure-dependent interfacial resistance to flow. The pressure
dependence of the resistance (R) to transport caused by interac-
tion of gas with the CNT can be expressed in terms of gas flux (J)
and the P via the relationship RG = P/JG. Since gas flux (JG) is
directly proportional to the ion flux via the relationship JG = KG·˚G,
the proportionality factor, KG, is then eliminated by normalizing
the flux data to the flux at one atmosphere total pressure. As such,
RG(normalized) = P/˚G.
Fig. 8 shows RG for the CNT membrane normalized to RG at 1
atmosphere total pressure. The data shown in Fig. 8 were obtained
over a wider range of pressures than the studies which compared
CNT and CT transmission characteristics.
Fig. 8 shows that flux resistance in the CNT introduc-
tion system increases with increasing pressure in the order
N2 > CH4 > O2 > Ar > CO2. This effect is consistent with gas interac-
tions with CNT walls becoming increasingly important at higher
pressures. The pore geometry of our CNTs is characterized by a
Knudsen number greater than 1, where the flow regime is dom-
inated by particle–surface collisions rather than particle–particle
collisions. The adsorption of gas molecules into the CNT channel
walls at increased pressures has been shown by others [12–14,27]
to lead to a deviation from pure Knudsen behavior and is hypoth-
esized to be the cause of decreased conductance with increasing
pressure.
4. Conclusions
The template method was used to fabricate CNT membranes.
Carbon was deposited within the pores of an AAO membrane
using the chemical vapor deposition process. The properties of
a CNT/AAO membrane directly coupled to the inlet system of a
mass spectrometer were compared to those of a conventional MS
introduction system, a direct-feed capillary tube (CT). The CNT
membrane exhibited fast mass-transport properties and enhanced
transmission of CH4. In the case of N2, the transmission prop-
erties of the CNT and CT introduction systems were broadly
comparable. From the work of other groups [12–14,27], it is rea-
sonable to conclude that interfacial resistance – interactions of
gas molecules with CNT walls – caused the conductance of the
CNT membrane to decrease with increasing pressure, progressively
favoring the transmission of heavier molecules. However, since
gas conductance through the CT introduction increases with pres-
sure in the order CO2 > Ar > O2 > CH4 > N2, comparison of CNT vs.
CT gas transmission always showed the following order (CNT/CT):
CH4 > N2 > O2 > Ar > CO2.
This work demonstrates that CNT membranes can be viable
introduction systems for mass spectrometers. Since CNT mem-
branes can be internally [28] and externally [29] functionalized,
CNTs have considerable promise for providing membrane systems
with a wide range of transmission characteristics. Future work will
focus on functionalized CNT membranes and their selective trans-
mission properties for various analytes in solution.
5. List of symbols
ID Raman D band intensity
IG Raman G band intensity
G gas x
˚ ion current
P partial pressure
S sensitivity factor
Q throughput
k throughput proportionality factor
C conductance
C conductance (normalized)
P total pressure gradient
mole fraction
R transport resistance
J ion flux
K ion flux proportionality factors
Acknowledgments
• Supported by the NSF Florida–Georgia Louis Stokes Alliance for
Minority Participation (FGLSAMP) Bridge to the Doctorate sup-
plements awards HRD #0217675 and GEO #0503536.
• Supported by the Alfred P. Sloan Minority Ph.D. Fellowship pro-
gram.
• The Office of Naval Research provided financial support through
Grant No. N00014-03-1-0479 and Contract No. N00014-07-C-
720.
• The assistance of Dave Edwards (USF-COT) is gratefully acknowl-
edged for SEM analyses and Ashok Kumar and his students (USF)
are gratefully acknowledged for assistance with Raman Spec-
trometry.
References
[1] S. Bauer, D. Solyom, Determination of volatile organic compounds at the parts
per trillion level in complex aqueous matrices using membrane introduction
mass spectrometry, Anal. Chem. 66 (1994) 4422.
[2] A. Keil, N. Talaty, C. Janfelt, R.J. Noll, L. Gao, Z. Ouyang, R.G. Cooks, Ambient
mass spectrometry with a handheld mass spectrometer at high pressure, Anal.
Chem. 79 (2007) 7734.
[3] S. Dheandhanoo, R.J. Ciotti, S.N. Ketkar, Atmospheric pressure sample inlet for
mass spectrometers, Rev. Sci. Instrum. 71 (2000) 4655.
[4] R.C. Johnson, R.G. Cooks, T.M. Allen, M.E. Cisper, P.H. Hemberger, Membrane
introduction mass spectrometry: trends and applications, Mass Spectrom. Rev.
19 (2000) 1.
6. L.D. Miranda et al. / Journal of Membrane Science 344 (2009) 26–31 31
[5] G. Hoch, B. Kok, A mass spectrometer inlet system for sampling gases dissolved
in liquid phases, Arch. Biochem. Biophys. 101 (1963) 160.
[6] K.H. Bennett, K.D. Cook, J.L. Falconer, R.D. Noble, Time-dependent perme-
ance of gas mixtures through zeolite membranes, Anal. Chem. 71 (1999)
1016.
[7] B.S. Ferreira, F. van Keulen, M.M.R. da Fonseca, On-line monitoring of dissolved
gases using microporous membrane inlet and mass spectrometry, Food Tech-
nol. Biotechnol. 38 (2000) 63.
[8] R.V. Parthasarathy, K.L.N. Phani, C.R. Martin, Template synthesis of graphitic
nanotubules, Adv. Mater. 7 (1995) 896.
[9] T. Kyotani, L. Tsai, A. Tomita, Preparation of ultrafine carbon tubes in nanochan-
nels of an anodic aluminum oxide film, Chem. Mater. 8 (1996) 2109.
[10] S.M. Cooper, B.A. Cruden, M. Meyyappan, R. Raju, S. Roy, Gas transport charac-
teristics through a carbon nanotubule, Nano Lett. 4 (2004) 377.
[11] D.A. Newsome, D.S. Sholl, Influences of interfacial resistances on gas transport
through carbon nanotube membranes, Nano Lett. 6 (2006) 2150.
[12] S. Kim, J.R. Jinschek, H. Chen, D.S. Sholl, E. Marand, Scalable fabrication of car-
bon nanotube/polymer nanocomposite membranes for high flux gas transport,
Nano Lett. 7 (2007) 2806.
[13] A. Noy, H.G. Park, F. Fornasiero, J.K. Holt, C.P. Grigoropoulos, O. Bakajin, Nanoflu-
idics in carbon nanotubes, Nanotoday 2 (2007) 22.
[14] J.K. Holt, H.G. Park, Y. Wang, M. Stadermann, A.B. Artyukhin, C.P. Grigoropou-
los, A. Noy, O. Bakajin, Fast mass transport through sub-2-nanometer carbon
nanotubes, Science 312 (2006) 1034.
[15] H. Masuda, M. Satoh, Fabrication of gold nanodot array using anodic porous
alumina as an evaporation mask, Jpn. J. Appl. Phys. 35 (1996) L126.
[16] G. Che, S.A. Miller, E.R. Fisher, C.R. Martin, An electrochemically driven actu-
ator based on a nanostructured carbon material, Anal. Chem. 71 (1999)
3187.
[17] P.P. Mardilovich, A.N. Govyadinov, N.I. Mukhurov, A.M. Rzhevskii, R. Paterson,
New and modified anodic alumina membranes. Part I. Thermotreatment of
anodic alumina membranes, J. Membr. Sci. 98 (1995) 131.
[18] D. Choi, S. Lee, C. Lee, P. Lee, J. Lee, K. Lee, H. Park, W. Hwang, Dependence of the
mechanical properties of nanohoneycomb structures on porosity, J. Micromech.
Microeng. 17 (2007) 501.
[19] D.J. Arrowsmith, A.W. Clifford, D.A. Moth, Fracture of anodic oxide formed on
aluminium in sulphuric acid, J. Mater. Sci. Lett. 5 (1986) 921.
[20] H. Hiura, T.W. Ebbesen, K. Tanigaki, H. Takahashi, Raman studies of carbon
nanotubes, Chem. Phys. Lett. 202 (1993) 509.
[21] Y.T. Lee, J. Park, Y.S. Choi, H. Ryu, H.J. Lee, Temperature-dependent growth of
vertically aligned carbon nanotubes in the range of 800–1100 ◦
C, J. Phys. Chem.
B 106 (2002) 7614.
[22] J.M. Holden, P. Zhou, X. Bi, P.C. Eklund, S. Bandow, R.A. Jishi, K.D. Chowdhury,
G. Dresselhaus, M.S. Dresselhaus, Raman scattering from nanoscale carbons
generated in a cobalt-catalyzed carbon plasma, Chem. Phys. Lett. 220 (1994)
186.
[23] I. Eswaramoorthi, L.-P. Hwang, Synthesis and characterisation of larger diam-
eter multi-walled carbon nanotubes over anodic titanium oxide template,
Letters to the Editor, Carbon 44 (2006) 2330.
[24] S. Welz, M.J. McNallan, Y. Gogotsi, Carbon structures in silicon carbide derived
carbon, J. Mater. Process. Technol. 179 (2006) 11.
[25] B.H. Kim, B.R. Kim, Y.G. Seo, A study on adsorption equilibrium for oxygen and
nitrogen into carbon nanotubes, Adsorption 11 (2005) 207.
[26] G. Arora, S.I. Sandler, Air separation by single wall carbon nanotubes: thermo-
dynamics and adsorptive selectivity, J. Chem. Phys. 123 (2005) 044705.
[27] A.I. Skoulidas, D.S. Sholl, J.K. Johnson, Adsorption and diffusion of carbon diox-
ide and nitrogen through single-walled carbon nanotube membranes, J. Chem.
Phys. 124 (2006) 054708.
[28] T. Kyotani, W. Xu, Y. Yokoyama, J. Inahara, H. Touhara, A. Tomita, Chemical
modification of carbon-coated anodic alumina films and their application to
membrane filter, J. Membr. Sci. 196 (2002) 231.
[29] K.K.S. Lau, J. Bico, K.B.K. Teo, M. Chhowalla, G.A.J. Amaratunga, W.I. Milne, G.H.
McKinley, K.K. Gleason, Superhydrophobic carbon nanotube forests, Nano Lett.
3 (2003) 1701.